Friday, November 4, 2016

WA: How Bad Is Barnett's Latest Polling?

I last covered Western Australian state polling in March, in an article called A New Species Of Strangeness In The West.  At that time, despite increasingly bad Newspolls for the Barnett Liberal government, there was a bizarre and very short-lived call to replace Opposition Leader Mark McGowan with former federal MP Stephen Smith.  Since then it is Barnett who has gone through the normal process of leadership speculation and challenge that strikes leaders who are actually polling badly.

It came to a head in mid-September when then Transport Minister Dean Nalder quit Cabinet and launched a leadership challenge, in which his proposal for a spill managed to get 15 votes out of 46 although there was almost no public support for him as leader (5.5% in ReachTEL) and a widespread perception that his challenge was hopeless from the start.  This would normally be the beginning of the end for an incumbent Premier, but the difference here is that if someone is going to throw Barnett under the bus, they will need to get a wriggle on.  It is just over four months til the election.

A new Newspoll gives contrasting results, with an improvement in the government's 2PP percentage from 46% in May to 48% now, but Barnett's worst set of personal ratings yet.  The relatively close 2PP has led to speculation that the government is back in the hunt.  The gap is closing, the election will be a cliffhanger, Barnett has clawed back support and Labor should be worried, and so on.

But really, there is not that much to see in the Newspoll 2PPs alone.  The primary vote changes consist of Labor dropping one point, the Greens dropping two, and Others gaining three (mainly on the back of the addition of One Nation).  On a 2PP basis this would normally be worth only about 1.2 points for the government, so it's quite likely that the 54 for Labor was 53-point-something and the 52 was 52-point-something, with the two-point shift being partly caused by rounding.

Even if the change was two whole points, that's not statistically significant and might be caused by sampling noise.  While it's better for the government to get 48 now rather than another 46, based on Newspoll alone its support level might not have even changed at all.

Polls Other Than Newspoll (Oh Yes They Do Exist)

Looking at other polls (the sort that aren't being mentioned much in discussion of the Newspoll), there was a 56-44 to Labor ReachTEL shortly after the Smith episode, which has since been followed by a 51-49 reading in September and a 52-48 result in October.  The ReachTELs are coming off higher primaries for the government (once undecided voters are removed, which the published detail annoyingly often fails to do) but they are also using respondent preferencing.  The 56 was more like a 53 or 54 and the more recent polls would point to a very close contest indeed if last-election preferences were used.  Generally, respondent preferences in polling tend to favour Labor, and to do so unduly even when there is a real shift in preferencing going on.  There is plenty of room for preferences to shift in WA though - the Coalition has recently got over 60% of Others preferences and about 30% of Greens preferences, both figures that could easily go down.

We've also had some Essential state readings (which had Labor up 53:47 in July-Sep after being up 51:49 in April-June) and the volatile and unreliable Morgan SMS series (52.5 to Labor in October, 51 to Labor in August, 51-49 to Liberal back in May).

Overall, everyone at the moment has Labor's lead somewhere around 52:48.  An aggregation would probably show the government recovering some ground since earlier this year, depending on how strongly it weighted the high-quality polls versus the less reliable ones.  But it hasn't been a big shift through the year, and while Labor's aggregated lead has never been overwhelming, it has also been persistent.

The Target Score

Much of the caution about a possible Labor victory concerns the need for a 10.1% uniform swing, which would mean Labor would need a 52.8% 2PP vote to win.  However, swings are never uniform and always vary from electorate to electorate, and this is why probability-based analysis is often useful to avoid misleading projections.  As I pointed out last time, when this is taken into account, the target swing drops to about 8.3% (51% 2PP), with the government's edge largely resulting from personal vote effects from the previous election.  But if the swing is not effectively random (it is regionally concentrated, or concentrated in seats that are more or less marginal, or concentrated in a given party's seats) then the target swing may be higher or lower than that.

From my reading of the comments of those who know the landscape well (like William Bowe) there is some reason to think the government might get smacked hard in volatile outer suburban target seats, so if anything the target statewide swing could be a bit lower and the 2PP playing field close enough to level.  A union-commissioned ReachTEL of the seat of Wanneroo (showing Labor ahead 57-43) should be treated with some caution because it appears to use respondent preferences, and because commissioned polls are often prone to selective release (if the results are bad for the commissioning source then you never see them). But probably not seven 2PP points worth of caution,

Government sources are being reported as expressing hope that they might hold on by "sandbagging" crucial marginals.  There are, however, many cases of this strategy being talked about in state elections, and few of it actually working.  The last Victorian election is a good example: the Opposition only led 52-48 in polling, but the government was unable to either get a swing back to it from that polling or to sandbag seats effectively.

The 2PP polling at the moment is far from showing the election is in the bag.  The best read of it at the moment seems to be that Labor would be likely to narrowly win. However, only a modest degree of polling error would be required for it to be, on paper, a very tight contest.  I would not place any faith in polling error from the Government's side.


The latest Newspoll shows Colin Barnett with a net satisfaction rating of -33 (28-61).  Only six Australian state Premiers have ever polled worse than this, and another two have equalled it.  Of the eight Premiers to poll a netsat of -33 or worse at any stage of their term, five jumped before they were pushed, Sir Joh Bjelke-Peterson was deposed by his own party, and Anna Bligh and Lara Giddings made it to the next election, where their parties were thrashed.  In fact, as documented in my earlier piece Unpopular State Premiers Have Dire Historic Fates, no Premier has ever recovered from worse than -24 at any stage of a term and gone on to win.  Barnett is nine points downwind from that and very close to the end of the term with little time to rebuild perceptions.  Moreover his bad ratings are not the result of a momentary scandal or unpopular tough policies, but are of the long-term, set-in, gradually-worsening kind.

When it comes to the better Premier question, Mark McGowan holds a 47-29 lead.  This frankly staggering eighteen-point margin equals the largest lead held by any state or federal opposition leader in a two-way poll since 1994.  (At one stage Bill Shorten held the same lead over Tony Abbott, who was deposed.)  As I noted in the previous piece, preferred-leader ratings can throw up some silly results early in a government's term, but at most times the skew to the incumbent leader is so severe that if a Premier is behind at all, then they are toast.

For whatever reasons (federal drag often being one of them) State Premiers do not have the kind of resilience to bad polling - be it 2PP, personal ratings or preferred leader scores - that Prime Ministers display.  Some of this is because when they poll very badly, they tend to resign or be removed.  We don't have that many data points for the performance of severely unpopular Premiers at elections for this reason, but the success rate even for Premiers who have been modestly unpopular at some stage of their term is dismal.

A little-mentioned aspect of Barnett's leadership is his age.  The other Premiers and Chief Ministers of today are in their forties except for Jay Weatherill who is in his early fifties; Barnett is 66.  Perhaps as 69 and 70 year old candidates battle for the Presidency of the United States we shouldn't read much into this, but Barnett would be the oldest Premier to be elected since Sir Joh Bjelke-Peterson in 1986.  Sir Joh's career would have ended at a younger age but for blatant malapportionment; before him there was Sir Charles Court in WA in 1980 and Sir Robert Askin in NSW in 1973.


In most regards, Barnett's Premiership fits the historic profile of one his party would have already ended.  This hasn't happened, whether because of his sheer dominance of the WA political scene for so long or because his party just plain doesn't have anyone else.  They are now in a quandary, because if they continue as they are defeat is likely, but they do not have time to install anyone else in time for a new leader to build their own brand.  Replacing Barnett would therefore be a gamble that the party's continued bad polling was mostly down to him and that voters would otherwise return his government even under a little-known new "nightwatchman Premier".  This was supposed to be the message of the infamous private poll (with a massive sample size) in August, but information on that poll's details is scanty, and polls that ask voters how they will vote under hypothetical circumstances are not very reliable.  Moreover, since August, crucial lead-in time for any new leader has gone to waste.

Against the picture of relatively close 2PP polling we have to set the big pictures that "disunity is death", that having the same party in power federally is a big drag on a state government that has been in power this long at the best of times, and that the Liberal Party seems to lack an adequate choice as leader.  There is not enough evidence to safely write the Barnett Government off, but if it does survive from here it will be a serious outlier in Australian state electoral history.

No comments:

Post a Comment

The comment system is unreliable. If you cannot submit comments you can email me a comment (via email link in profile) - email must be entitled: Comment for publication, followed by the name of the article you wish to comment on. Comments are accepted in full or not at all. Comments will be published under the name the email is sent from unless an alias is clearly requested and stated. If you submit a comment which is not accepted within a few days you can also email me and I will check if it has been received.